Hi, 2016-01-07 12:48 GMT+01:00 foo86 <fooba...@gmail.com>: > bench dca pcm_f32le > bench dca2 pcm_f32le > bench libdcadec pcm_s32le
OK, that was mostly out of curiosity, as "dca2" has benefits surpassing such issues anyway. And the improvement is 10% for where it matters (raspberry). > synth_filter integrates just nicely without any changes, there is no need for > rewrite. There is 64 subband version of it needed though, as well as fixed > point versions for 32 and 64 subbands. In my recollection, the code was weirdly shaped, so I could have imagined needs for change (e.g., your libdcadec may not be able to reuse it). > Other parts can't be directly reused unfortunately. They need to be converted > to operate on integer input and process entire frame at once (not just one > subsubframe). The gap between dca and dca2 will probably increase, as these "constrains" should actually help make the operation faster. > Temporary converting the input to floating point before calling them (what > current dca code does) seems to negate any performance gain from using these > assembly optimized routines in the first place... I think it is ok to change the code and remove the arch-specific dsp implementations. Whether to put that specific removal in a separate patch is up to you. -- Christophe _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel